A Multi-Model Study of the Association between BMI and AMH in Chinese Women

OBJECTIVE: To determine whether BMI is associated with AMH in Chinese women, and whether PCOS, age, or testosterone modify this association. METHODS: We analyzed 670 women (20-39 years) from Hangzhou Women's Hospital. Primary analyses included linear regression and generalized additive models (GAM) with thin-plate splines, and multi-group structural equation modelling to test PCOS effect modification. Secondary analyses included interaction tests and Chinese BMI category comparisons. All models adjusted for age, FSH, LH, and testosterone. RESULTS: BMI showed no significant linear association with log-AMH (β = 0.008; adjusted R2 = 0.398). GAMs confirmed a near-linear relationship without threshold effects. However, underweight women (BMI < 18.5 kg/m2) had significantly lower log-AMH than normal-weight women (β = -0.14, P = 0.036), with no significant differences for overweight or obese categories. PCOS did not modify the BMI-AMH association. CONCLUSIONS: BMI is not a clinically meaningful predictor of ovarian reserve as a continuous variable. Lower AMH in underweight women highlights a subgroup warranting attention in reproductive assessments, while PCOS status does not influence this relationship.